With each successive era of LLMs, from early scaling, to reasoning, to agentic, open models have taken half as long to catch up to the first closed model

Comparing open vs. closed models across the eras of frontier models, Is the gap narrowing?